[Eeglablist] How many dipoles to fit?

[Scott Makeig]

A few more thoughts on one- vs. two-dipole fitting:

A general two-equivalent-dipole model can always fit a given (IC or other)
scalp map better than a one-dipole model, but in the case of a very-near
dipolar source reflecting synchronous activity across a cortical patch, the
second dipole fit may only reduce errors produced by the head model, etc...
There may also be (incorrect) local minima in the 'error surface' for
general two-dipole fits.

For symmetric two-dipole fitting, the chance of false local minima is low to
none and the two-dipole solution might not reduce the residual error beyond
the one-dipole model in some instances. However, there is a chance that the
second dipole might just be used to correct for head model or numeric error,
and not reflect physiological reality (cortical synchrony).

As Arno responded, in many instances an IC map with two bilaterally
symmetric foci clearly calls for a symmetric two-dipole model fit. Other
instances are less clear, however, and so far we have not worked on
a automated test for  deciding to fit one versus two dipoles for a given IC.
One of the SCCN group and/or some other researchers using ICA may be
interested in working on that.

-Scott Makeig

On Sun, Jun 5, 2011 at 9:31 AM, Arnaud Delorme <arno at ucsd.edu> wrote:

> Dear Igor,
>
> we fit 95%-99% of the components using a single equivalent dipole. We only
> fit 2 symmetrical dipoles for bilateral occipital alpha ICA components which
> are easily recognizable. Note that a large number of components cannot be
> fitted well with a single dipole (or 2 bilateral dipoles). These include
> artifactual components, or components that do not account for much of the
> data variance and reflect noise in the EEG data or in the ICA algorithm
> optimization procedure.
>
> Best regards,
>
> Arno
>
> On May 27, 2011, at 6:06 AM, Igor Riecansky wrote:
>
>
> Dear EEGLAB community,
>
> I am starting to use DIPFIT2 to localize brain sources of PCA/ICA EEG
> components. I face a problem to decide how many dipoles to fit, i.e. one
> unilateral vs two bilateral symmetrical. Some papers simply always assume 2
> symmetrical sources. However, if component's topography is highly
> asymmetric, this seems not to be justified. Furthermore, in case of a
> symmetric topography one could often reasonably assume both one unilateral
> or two bilateral sources both originating from brain structures close to the
> midline. Of course, solutions may substantially differ for one and two
> sources. What is you experience and practice? How do you solve this problem?
> Do you use any quantitative test? Residual variance is always lower when two
> sources are fitted compared to one source. Is there some rule to consider
> this reduction as significant, or can the reduction of residual variance be
> statistically tested?
>
> I will greatly appreciate any feedback and help. Thanks.
>
>
> Igor
>
> -----------------------------
> Igor Riecansky, MD PhD
> SCAN-Unit
> Institute of Clinical, Biological and Differential Psychology
> Faculty of Psychology
> University of Vienna
> Liebiggasse 5, A-1010 Austria
> Tel.: +43-1-4277-47 509
> Room No. O3.28
> E-mail: igor.riecansky at univie.ac.at
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>
>
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>



-- 
Scott Makeig, Research Scientist and Director, Swartz Center for
Computational Neuroscience, Institute for Neural Computation & Adj. Prof. of
Neurosciences, University of California San Diego, La Jolla CA 92093-0559,
http://sccn.ucsd.edu/~scott
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20110605/e7a64858/attachment.html>